FIELD OF THE INVENTION

[0001]
The present invention relates to combinatorial and sequential logic circuits, and more particularly to combinatorial and sequential logic circuits operating at a fractional clock rate.
BACKGROUND OF THE INVENTION

[0002]
Combinatorial logic devices produce one or more output signals in response to logical combinations of one or more input signals. Thus, combinatorial networks can function to indicate the presence of a given set of input signals by producing a corresponding output signal. Decoders, adders, logic gates and shifters are examples of combinatorial logic networks. A simple exemplary decoder produces an output signal indicating the state of two input variables. Since there are four possible combinations of two binary input variables, the input state can be determined by a decoder capable of producing one of four output signals. The output signal thus indicates or decodes the input state. Combinatorial networks are conventionally built from basic logic circuits including AND gates, OR gates, inverters, etc.

[0003]
Sequential logic circuits (or register logic circuits) produce an output signal in response to one or more input signals and a clock trigger signal. These synchronous circuits are typically constructed from basic flipflops, that change state only at the active edge of a clock signal (where the active edge can be the leading clock edge or the lagging clock edge). Synchronous flipflops and registers are examples of sequential networks. The clock signal synchronizes the operation of sequential networks by triggering registers and latches and advancing flipflops.

[0004]
Most conventional digital systems are designed to operate at a socalled full clock rate equal to the data rate. These synchronous circuits are triggered by the active edge of the clock signal. Each synchronous circuit has a known set up and hold time, representing the time during which the input signal must remain constant, both before and after the active clock edge, so that the circuit can perform its intended operation in response to the input signal. However, clock speeds are ever increasing in response to requirements for faster processing of digital data. As the period of each clock cycle decreases, there may be insufficient time for a circuit element (such as a gate) to perform its function during the interval between the active edge of two successive clock cycles. That is, when the clock period is less than the combined set up and hold time of a synchronous circuit element the output signal maybe erroneous.

[0005]
To overcome this difficulty, circuit designers convert the fullrate digital design to a halfrate clock design. The halfrate design runs at half the original clock frequency (thus the clock period is doubled), but performs twice the number of data processing steps during each clock cycle, so that the output signals are equivalent to and occur at the same time as the fullrate design. To perform at the doubled rate, the halfrate design typically includes twice the number of circuit elements as the fullrate circuit, including both the registered or sequential logic elements and the nonregistered or combinatorial logic elements.

[0006]
The conversion process from the full rate to the halfrate design can be a time consuming task. Control circuits or other circuits having a considerable number of fullrate feedback paths can be difficult to convert to a halfrate system. For example, converting a fullrate state machine to halfrate operation is a complex and timeconsuming process, since there are a plurality of states and each state has several feedforward paths. Doubling the circuit elements so that operation at halfrate produces output signals that are timecoincident with a full rate clock must account for these various states and their feed forward paths to the next state.

[0007]
The conversion process is essentially ad hoc, relying primarily on the skill level and knowledge of the circuit designer. Thus the results of the conversion process often vary considerably in terms of the circuit area consumed on an integrated circuit device in which the design is implemented, power consumption, and overall effectiveness of the halfrate design when compared with its fullrate counterpart.
BRIEF SUMMARY OF THE INVENTION

[0008]
The present invention describes a fractionalrate clocked logic circuit or logic block and a method for converting a fullrate clocked logic circuit to a fractionalrate clocked logic circuit. The fullrate clocked logic circuit comprises combinatorial functions and sequential functions for operating on block input signals to produce block output signals, wherein the sequential functions operate at a fullrate clock frequency. The method comprises deriving combinatorial logic elements based on the combinatorial functions, wherein each of the combinatorial logic elements is responsive to a subset of the block input signals, for producing combinatorial signals. The method further comprises deriving sequential logic elements based on the sequential functions, where each of the sequential logic elements is responsive to a subset of the combinatorial signals and a fractional clock signal, for producing register signals. A frequency of the fractional clock signal is a fraction of the fullrate clock frequency. At least one of the sequential logic elements further produces a register feedback signal. Certain of the combinatorial logic elements are further responsive to a subset of the combinatorial signals, and at least one of the combinatorial logic elements is further responsive to the register feedback signal. The register signals are combined for producing the block output signals.

[0009]
The logic block processes input signals for producing output signals. The block comprises a deinterleaver for receiving and deinterleaving the input signals into a plurality of input signal groups and a plurality of combinatorial function elements each one for receiving one of the plurality of input signal groups for producing a combinatorial signal. A plurality of sequential function elements each receive a combinatorial signal group and a clock signal for producing a register signal. At least one of the plurality of sequential function elements produces a register feedback signal. Each one of the plurality of combinatorial function elements further receives one of the combinatorial signals, and at least one of the plurality of combinatorial function elements receives the register feedback signal. An interleaver is responsive to the register signals for producing the output signals.
BRIEF DESCRIPTION OF THE DRAWINGS

[0010]
The present invention will be more easily understood and the advantages and uses thereof more readily apparent when considered in view of the following detailed description of the invention when read in conjunction with the following figures wherein:

[0011]
FIG. 1 is a block diagram of a prior art fullrate clock digital circuit including combinatorial and register circuit functions;

[0012]
FIG. 2 is a block diagram of a halfrate clock digital circuit implementing the function of the fullrate clock digital circuit of FIG. 1 according to the teachings of the present invention;

[0013]
FIG. 3 is a timing diagram illustrating certain operating principles of the halfrate clock digital circuit of FIG. 2;

[0014]
FIG. 4 is a block diagram of a cascaded halfrate clock digital circuit;

[0015]
FIG. 5 is a block diagram of an exemplary prior art fullrate clock digital circuit;

[0016]
FIG. 6 is a block diagram of an exemplary halfrate clock digital circuit according to the teachings of the present invention;

[0017]
FIG. 7 is a block diagram of second embodiment of a halfrate clock digital circuit according to the teachings of the present invention; and

[0018]
FIG. 8 is a block diagram of a fractional rate clock digital circuit according to the teachings of the present invention.

[0019]
In accordance with common practice, the various features of the present invention are not drawn to scale, but are drawn to emphasize specific features relevant to the invention. Reference characters denote like elements throughout the figures and text.
DETAILED DESCRIPTION OF THE INVENTION

[0020]
Before describing in detail the particular halfrate clock digital circuitry and the method for converting a fullrate clock digital circuit to a halfrate clock circuit in accordance with the present invention, it should be observed that the present invention resides primarily in a novel and nonobvious combination of elements and method steps. Accordingly, these elements and steps have been represented by conventional elements in the drawings, showing only those specific details that are pertinent to the present invention so as not to obscure the description with structural details that will be readily apparent to those skilled in the art having the benefit of the description herein.

[0021]
FIG. 1 illustrates a prior art fullrate clock logic block 8 comprising a combinatorial function 12 and a register function 14 (also referred to as a sequential function). A plurality of m+1 logic signals (designated 0, 1, 2 . . . m) comprise a signal vector X (designated X[m:0]), input to the combinatorial function 12 and processed in conjunction with a signal vector of Y logic signals according to a defining function F(X, Y) to produce an output signal vector Z[n:0] that is in turn provided as an input to the register function 14. Note X designates one or more logic signals or bits originating from a prior circuit element, and Y designates one or more logic signals or bits fed back from another circuit element (such as the register function 14).

[0022]
The register function 14 is responsive to a fullrate clock signal 16 at a clock terminal 18 for controlling the synchronous operation thereof. The register function 14 provides both a register feedback vector (designated Y[n:0]) and a block output vector (designated Y[j:0]), which is a subset of the register feedback vector. The register feedback vector Y[n:0] is provided as a feedback input to the combinatorial function 12 as described to above.

[0023]
As applied to digital systems, the combinatorial function 12 represents one or more combinatorial digital functions, for example, a multiplier, shifter, adder, state encoder, or any combination of such elements, and is described by the function F(X,Y) operating on the two input vectors, the block input vector X[m:0] and the register feedback vector Y[n:0]. Likewise, the register function 14 represents one or more known synchronous circuit elements. By appropriately segregating the combinatorial and the sequential circuit elements of a digital system, most digital systems can be represented as a cascade of the fullrate clock logic block 8 of FIG. 1. In the cascade configuration the register block output vectors Y[j:0] of block “p” are provided as inputs to a block “p+1”. Typically, the combinatorial functions of each cascaded block are implement different logic functions.

[0024]
FIG. 2 illustrates a halfrate clock logic circuit or block constructed according to the teachings of the present invention, comprising combinatorial functions 30 and 32 and register functions 34 and 36. In converting the fullrate clock design of FIG. 1 to the halfrate clock design of FIG. 2, according to the teachings of the present invention the combinatorial function 12 of FIG. 1 is duplicated and represented by the combinatorial functions 30 and 32 of FIG. 2. That is, the function F(X,Y) for the combinatorial function 12 is implemented in both the combinatorial functions 30 and 32. The register functions 36 in the halfrate design is identical to the register function 14 in the fullrate design. The register function 34 is identical to the register function 14 with the exception that the register function 34 excludes the registers that generate feedback signals.

[0025]
With reference to FIG. 1, the serial stream of block input vectors X[m:0] is segregated (deinterleaved) into X_even and X_odd vectors (or signal groups) by a deinterleaver 37. With reference to FIG. 3, the deinterleaver 37 receives one X[m:0] block input vector (for example, X_{0}, X_{1}, X_{2}, X_{3}) during successive fullrate clock periods, where the fullrate clock frequency is designated f_{c }and thus the fullrate clock period is 1/f_{c }as labeled on the time axis. Two consecutive block vectors X_{0}, X_{1 }are output from the deinterleaver 37 during a single halfrate clock duration 38, where a halfrate clock cycle duration is equivalent to two full rate clock cycles as shown. X_{0}, received first in time, is identified as the X_even block vector, and X_{1}, the block vector received second in time, is identified as the X_odd block vector. During the next two halfrate clock cycles the block vectors X_{2}, X_{3 }are output from the deinterleaver 37 as shown.

[0026]
Returning to FIG. 2, the X_even[m:0] vectors are supplied as inputs to the combinatorial function 30. The odd vectors, designated X_odd[m:0], are supplied as inputs to the combinatorial function 32. The combinatorial function 30 produces Z_even[n:0] output vectors that are input to the register function 34 and to the combinatorial function 32. The combinatorial function 32 produces Z_odd [n:0] output vectors that are provided as inputs to the register function 36.

[0027]
The register function 34 provides even block output vectors Y_even[j:0]. Odd block output vectors Y_odd[j:0] are produced by the register function 36. The combination of the even and the odd block output vectors is equivalent to the block output vectors produced by the register function 14 in the fullrate clock circuit of FIG. 1. Additionally, the register function 36 produces odd register feedback vectors Y_odd[n:0] that are supplied as input signals to the combinatorial function 30 for operation according to the function F(X,Y). The odd register feedback vectors Y_odd[n:0] include all output vectors from the register function 36, while the odd block output vectors Y_odd[j:0] are a subset of the feedback signals. As can be seen, a halfrate clock signal, operating at a frequency of onehalf the fullrate clock is supplied as a clocking input to the clock input terminals 43 and 44 of the register functions 34 and 36 respectively.

[0028]
The Y_even and Y_odd block output vectors from the register functions 34 and 36, respectively, can be combined by interleaving (that is, by selecting the first Yeven output vector then the first Y_odd output vector, followed by the second Y_even output vector, etc.) in an interleaver 46 to produce the same composite block output vectors as provided by the fullrate clock register function 14 of FIG. 1. Note that the halfrate clock outputs are produced two at a time during each halfrate cycle (prior to interleaving), as compared to one at a time during each fullrate clock cycle of the FIG. 1 prior art circuit.

[0029]
According to the teachings of the present invention, a fullrate clock circuit, such as that illustrated in FIG. 1, can be converted to a halfrate clock circuit, such as that illustrated in FIG. 2, according to the following steps.

[0030]
1. Identify the block input signal vectors (X[m:0]), register input signal vectors (Z[n:0]), register feedback signal vectors (Y[n:0]), and block output signal vectors (Y[j:0]).

[0031]
2. Select alternating block input signals from the fullrate clock design, designated X_even[m:0], and provide them as inputs to a first (even) combinatorial function (such as the combinatorial function 30 of FIG. 2), which has the same controlling function (F(X,Y)) as the combinatorial function of the fullrate design.

[0032]
3. Provide the alternating fullrate block input signal vectors, designated X_odd[m:0], as input signals to a second (odd) combinatorial function, such as the combinatorial function 32 of FIG. 2, which implements the same function F(X,Y) as the fullrate clock combinatorial function. When considered with reference to the serial block input signal vectors of the fullrate design, each of the X_odd input signal vectors is sequentially later in time than the corresponding X_even input signal vectors.

[0033]
4. Supply the combinatorial output vectors Z_even[n:0] from the first (even) combinatorial function as input signal vectors to a first (even) register function, such as the register function 34.

[0034]
5. Supply the combinatorial output signal vectors from the second (odd) combinatorial function (Z_odd[n:0]) as inputs to a second (odd) register function, such as the register function 36.

[0035]
6. Supply the Y_odd[n:0] register feedback signal vectors (from the second (odd) register function) as inputs to the first (even) combinatorial function.

[0036]
7. Supply the Z_even[n:0] combinatorial output signal vectors from the first (even) combinatorial function as inputs to the second (odd) combinatorial function.

[0037]
8. Combine, by interleaving, the Y_even[j:0] and the Y_odd[j:0] block output signal vectors from the register functions to produce a composite Y[j:0] output signal vector, which is identical to the output signal vector from the register function of the fullrate circuit.

[0038]
The abovedescribed procedure for converting from a fullrate clock to a halfrate clock digital logic circuit can be used in conjunction with digital circuitry sharing a common clock. The resulting halfrate circuit reduces power consumption.

[0039]
Use of the halfrate clock circuit can also improve throughput if the clock frequency is increased from the design value, as the fullrate circuit will fail to meet the timing requirements of an increased clock rate before the halfrate clock circuit will fail. That is, the halfrate circuit has the potential to provide more throughput because the halfrate clock can be operated faster than onehalf of the maximum fullrate clock frequency without exceeding the timing limitations of the register elements. In the embodiment where the halfrate clock frequency is exactly onehalf the fullrate clock frequency, both circuits have the same throughput. This is possible because in the halfrate circuit, although the duration of the combinatorial operations is doubled as the output vectors from one combinatorial function (the combinatorial function 30) are supplied as inputs to the other combinatorial function (the combinatorial function 32), the register setup and hold times (the registers 34 and 36) are not doubled. For example, assume the combinatorial function of a fullrate circuit takes 8 ns from input to output, and the register functions consume 2 ns for the set up and hold operations. Thus the full rate operation speed is limited to 8+2=10 ns. If the circuit is operated at this maximum possible speed a new output signal vector is produced every 10 ns. For the halfrate circuit, the two combinatorial functions operate in series, and assuming the same setup and hold times for the register functions, two halfrate outputs are produced every 8+8+2=18 ns, or one output signal vector every 9 ns. Thus, the halfrate circuit exhibits better circuit speed than the fullrate case. In practice, further improvements can be realized by combining and optimizing the two even and odd combinatorial functions 30 and 32 using available synthesis tools. The resulting optimized combinatorial functions would be expected to operate faster than combined the 8+8=16 ns.

[0040]
As can be seen from FIG. 2, the register function 34 provides only the even block output signal vectors Y_even[j:0], whereas the register function 36 provides both odd register feedback signal vectors (Y_odd[n:0]) and odd register block output signal vectors (Y_odd[j:0]). Typically the number of register block output signals is considerably less than the number of register feedback signals. For example, for a block comprising a 10bit shift register, there are nine feedback signals but only one output vector.

[0041]
The fact that the even register function 34 does not supply feedback vectors to other elements of the block can be advantageously employed in a cascaded logic circuit design comprising several stages of odd and even combinatorial and register functions. Such an embodiment is illustrated in FIG. 4, where even and odd block input signal vectors (X_even[m:0] and X_odd[m:0]) are provided to the combinatorial functions 30 and 32 as in FIG. 2. Also, the feedback signal vectors provided as inputs to the combinatorial functions 30 and 32 are identical to the feedback signal vectors of the FIG. 2 embodiment.

[0042]
In the cascaded design of the FIG. 4, even combinatorial output signal vectors from the combinatorial function 30 are supplied directly as input vectors to an odd combinatorial function 50. Thus the register function 34 of FIG. 2 is not present, saving power and reducing device size.

[0043]
Continuing with the FIG. 4 cascaded embodiment, the odd register output vectors from the register function 36 are provided as combinatorial inputs to an even combinatorial function 52. The even combinatorial output vectors therefrom are supplied as inputs to the combinatorial function 50 and are also supplied as inputs to the next odd combinatorial function in the cascade.

[0044]
The odd combinatorial output vectors from the odd combinatorial function 50 are supplied as input vectors to a register function 60, for producing odd register feedback vectors that are provided as an input to the combinatorial function 52, and for producing odd register output vectors that are supplied as input vectors to the next even combinatorial function (not shown) in the cascade chain. Thus to continue with the cascade of circuit elements, additional halfrate clock blocks (comprising an even and an odd combinatorial function and an odd register function) can be added to the cascaded design of FIG. 4. In the circuit illustrated in FIG. 4, since a halfrate register is not present in the first stage, the halfrate clock design has the same number of register elements as the fullrate clock design, saving both circuit area and power.

[0045]
Note also that the FIG. 3 cascade structure adds no additional latency when the halfrate blocks are combined. Typically, according to the prior art, when halfrate blocks are cascaded each stage adds one fullrate clock cycle of latency when compared with the fullrate structure.

[0046]
Recent studies of highspeed circuit designs have suggested that the power dissipation of the clock network in some cases is approaching 50 percent of the total circuit power consumption. According to the teachings of the present invention, the clock power consumption is reduced by a factor of two, since the clock is driving the same number of circuit elements at half the original clock frequency.

[0047]
FIGS. 5 and 6 present an example of a fullrate clock circuit and a corresponding halfrate clock circuit constructed according to the teachings of the present invention. Assume the requirement is for a simple integrator where, new_value=old_value+input value. The addition will be performed in an adder 80 (see FIG. 5), which is one example of a combinatorial function, such as the combinatorial function 30 of FIG. 2. The adder 80 performs the arithmetic addition operation on two input values (X,Y) and produces an output value Z. In a sequential transfer circuit 82, (which is one example of a register function, such as the register function 34 of FIG. 2), the input value (Z) is transferred to a register whose outputs are designated Y, and Y is fed back to the adder 80 since Y now represents the cumulative sum. Operation of the sequential transfer circuit occurs at the active edge of a fullrate clock signal input to the sequential transfer circuit 82. Thus the adder 80 and the sequential transfer circuit 82 implement the following operations.
Z=X+Y
Y←Z (at the next clock edge)

[0048]
Let X_{0 }and X_{1 }represent two sequential input vectors to the FIG. 5 circuit. When X_{0 }is present at a first input terminal of the adder 80, the previous value of Y (designated Y_{−1}) representing the cumulative sum to that point, is present at a second input terminal thereof. Thus the adder output is:
Z _{0} =X _{0} +Y _{−1}.
Z_{0}, which represents the new cumulative sum, is supplied as input to the sequential transfer circuit 82, the output of which is
Y_{0}←Z_{0}.

[0050]
Z_{0}=Y_{0 }is also input to the adder 80 for summing with the next input value, X_{1}, which is now present at one input terminal of the adder 80. The result is
Z _{1} =X _{1} +Y _{0}, and
Y_{1}←Z_{1 }

[0051]
Thus in two successive fullrate clock cycles, two additions are performed, producing two cumulative sum output values (Y_{0 }and Y_{1}, wherein Y_{1 }is the most recent cumulative sum) from the sequential transfer circuit 82.

[0052]
FIG. 6 illustrates the same operations as FIG. 5 using a halfrate clock circuit constructed according to the teachings of the present invention. Even adder 90 and odd adder 92 are functionally identical to the adder 80 of FIG. 5. Even and odd sequential transfer circuits 94 and 96, respectively, are functionally identical to the sequential transfer circuit 82 of FIG. 5, and both are responsive to a halfrate clock signal. The output vector from the even adder 90 is provided as an input to the odd adder 92 and to the even sequential transfer circuit 94. The output vector from the odd adder 92 is provided as an input to the odd sequential transfer circuit 96. A register feedback vector from the odd sequential transfer circuit 96 is input to the even adder 90. Block output vectors are taken from the block outputs of the even and odd sequential transfer circuits 94 and 96, respectively.

[0053]
Two successive input vectors, designated X
_{even }and X
_{odd}, or consistent with the nomenclature of
FIG. 5, referred to as X
_{0 }and X
_{1}, are provided as inputs to the even and odd adders
90 and
92, respectively, through the deinterleaver
37. The output vectors are as follows:

 From the even adder 90: Z_{0}=X_{0}+Y_{−1 }
 From the odd adder 92: Z_{1}=X_{1}+Z_{0 }
The output vectors from the even and odd sequential transfer circuits 94 and 96 are:
 From the odd sequential transfer circuit 94: Y_{0}←Z_{0 }
 From the even sequential transfer circuit 96: Y_{1}←Z_{1 }
Interleaving the Y_{0 }and Y_{1 }vectors (taking Y_{0 }first) yields the same result as the full rate clock circuit of FIG. 4, that is, Y_{0 }is the first cumulative sum and Y_{1 }is the second, or more recent, cumulative sum.

[0060]
FIG. 7 is a block diagram of an alternative halfrate clock digital circuit according to the teachings of the present invention. In the FIG. 7 embodiment, the even register feedback vectors are supplied as inputs to the combinatorial function 32. The output vectors from the combinatorial function 32 (Z_odd[n:0]) are input to the combinatorial function 30. The FIG. 7 embodiment operates similarly to the FIG. 2 embodiment, but the combinatorial function input vectors and the register feedback vectors are reversed in the FIG. 7 embodiment. The FIG. 7 embodiment can also be employed in a cascaded system similar to that illustrated in FIG. 4.

[0061]
The method and apparatus of the present invention can also be extended to fractional clock rates other than the onehalf clock rate described above. In an embodiment of a onethird clock rate circuit, illustrated in FIG. 8, a deinterleaver 100, in response to three consecutive full rate vectors X_{0}, X_{1 }and X_{2 }produces three simultaneous output vectors X_{0}, X_{1 }and X_{2 }that are supplied as inputs to identical combinatorial functions 104, 106 and 108 at onethird the fullrate clock. The output vector Z_{0 }from the combinatorial function 104 is supplied as an input to the combinatorial function 106; the output vector Z_{1 }from the combinatorial function 106 is supplied as an input to the combinatorial function 108 as shown. The output vectors Z_{0}, Z_{1 }and Z_{2 }are also supplied as inputs to register functions 114, 116 and 118, for producing block output vectors Y_{0}, Y_{1}, and Y_{2}. The register functions 114, 116 and 118 are responsive to a onethird clock rate signal. The register feedback vector Y_{2}′ is fed back to the combinatorial function 104. An interleaver 120 produces a fullrate vector Y in response to the onethird rate vectors Y_{0}, Y_{1}, and Y_{2}.

[0062]
While the invention has been described with reference to preferred embodiments, it will be understood by those skilled in the art that various changes may be made and equivalent elements may be substituted for elements thereof without departing from the scope of the present invention. Further, the scope may include any combination of elements from the various embodiments set forth herein. In addition, modifications may be made to adapt a particular situation to the teachings of the present invention without departing from its essential scope. Therefore, it is intended that the invention not be limited to the particular embodiment disclosed as the best mode contemplated for carrying out this invention, but that the invention will include all embodiments falling within the scope of the appended claims.